A hybrid localization algorithm of RSS and TOA based on an ensembled neural network

In a localization system, time of arrival (TOA) and receive signal strength (RSS) technique are widely used to estimate the location of a mobile station (MS) in many sensor networks. To enhance the performance of MS location estimation accuracy, a new ensembled neural network (NN) is employed with the hybrid TOA and RSS measurements. Firstly, the basic least squared based method for the localization is derived, which can be used for obtaining the accurate output for the neural network training. Then TOA and RSS measurements are respectively preprocessed and employed to train the individual Back propagation (BP) neural network. At last, the weighted average method is presented to design total results based on all the individual neural network, in which the weights are adaptively designed by the training errors. The proposed hybrid method is compared with the single method and its Cramer-Rao lower bound(CRLB) in the simulation, which validate that the proposed algorithm has better efficiency and performance for localization than the other methods.